Project Summary/Abstract Exploiting new sources of information in density-modification and model-bullding This is a project to develop algorithms for constructing atomic models representing macromolecular crystal structures at resolutions worse than 3 - 3.5 A, for combining the power of density modification and refinement, and for using all the powerful tools in the Phenix environment to automate the structure determination process. Success in this project will make structure determination more robust and more rapid, lead to more accurate models, and allow structure determination to be carried out with lower-resolution data. The project will be carried out in close collaboration with the Berkeley, Duke and Cambridge groups. The increased power of the methods developed will allow automation of structure determination at lower resolutions than is currently feasible, saving large amounts of time and money for structural biologists around the world in academia, government and industry, and allowing levels of error analysis and validation that are completely impractical when structure determination is done manually. During the past 5 years of the Phenix project we have worked with the Berkeley, Duke and Cambridge groups to develop fully automated procedures for structure determination X-ray crystallography with any combination of MAD, SAD, and MIR methods. As a foundation for these high-level procedures, we have developed new methods for evaluation of the quality of experimental electron density maps, for identification of noncrystallographic symmetry from density maps, for multi-crystal averaging, and for identification of unknown ligands, and a suite of new tools for rapid model-building. Further we have embarked on an exciting collaboration with Cambridge group and the Baker laboratory to combine algorithms from the structure modeling field with complementary algorithms from the crystallography field. We plan now to focus our efforts on developing methods that will accelerate the determination of structures that remain challenging today. These challenging structures include those with data extending only to resolutions of 3.0 - 3.5 A or worse, those with weak experimental phase information, and those with molecular replacement information from distant homology models. To achieve this goal, we propose to work with the Berkeley, Cambridge and Duke groups to develop resolution-dependent model-building tools, to combine crystallographic pattern-based model-building with complementary energy-based modeling tools from the structure-modeling field, and to develop a framework that will merge density modification and refinement into a single more powerful procedure. In parallel with these new efforts we will continue our role in the development of an integrated, flexible, and easy-to-use Phenix system for macromolecular structure determination. This work will allow structural biologists to determine 3-dimensional structures of macromolecules that are of key importance in understanding disease and maintaining human health.